Building humanoid robots for the home: 1X's decade-long bet

Executive overview

Physical labor is becoming the next scarce resource — not because demand is falling, but because birth rates are declining and costs are rising. 1X is building humanoid robots for the home to address this, using a fundamentally different design paradigm from industrial robotics.

The key insight is that robots living among humans must be low-energy, compliant, and socially aware — not precise and powerful. Ten years of proprietary tendon-drive technology and vertically integrated manufacturing underpin a consumer-first go-to-market strategy.

Physical abundance through intelligent machines requires robots that can safely learn by doing, not just execute pre-programmed tasks.

Why classical robotics fails in the home

• Industrial robots use ~100:1 gear ratios — internal components spin at 20,000 RPM during normal movement
• High rotational energy makes safe collisions impossible outside a calibrated factory environment
• Factory robots work by stopping just before contact; homes offer no such precision
• Honda Asimo failed because it relied on environmental assumptions that break down in real-world chaos
• Safe home robots need low kinetic energy, compliant joints, and soft materials — humans are the model

1X's tendon-drive approach

• Tendon-drive systems (cable-driven actuators) minimize energy storage in moving limbs
• No off-the-shelf components exist — 1X builds custom motors, actuators, and the machines that make them
• Vertical integration extends to factory automation equipment
• Simplification target: reduce complexity from car-level to household-appliance level
• Low part count, light materials, no special alloys — designed for manufacturability from the start

How NEO learns

• Training starts with internet data, then synthetic simulation, then real robot data
• Teleoperation bridges the gap: a human operator embodies the robot, transferring knowledge directly
• Autonomous capability bootstraps from teleoperation data, then improves through real-world iteration
• Any household task is inherently social — navigating a kitchen requires communicating intent, not just physical dexterity
• Intelligence scales with diverse, real-world experience; lab-only training doesn't generalize

Consumer-first go-to-market

• Enterprise adoption of genuinely new technology is structurally slow: IT gatekeeping, labor unions, risk-averse leadership
• Consumer adoption creates bottom-up pressure that eventually forces enterprise uptake — ChatGPT is the template
• Early home deployments build the real-world dataset needed to improve the product
• Deep tech companies must generate revenue during the journey, not just at launch

Founder lessons from 10 years in hard tech

• Failure is only unacceptable if you didn't fully try or didn't extract the lesson
• Culture that tolerates failure under pressure is the actual test — easy to state, hard to maintain on deadlines
• Letting go of team members who had relocated to Norway during COVID was the most painful decision
• Picking a problem that excites people is the best recruiting advantage in deep tech
• The grind is real; enjoyment of the work is what makes it sustainable